low birthweight: (1) Are neighborhood- .... TABLE 1-Variables Used in Multilevel Analysis of Risk for Low Birthweight, ..... Sampson R. Linking the micro- and.
Neighborhood Risk Factors for Low Birthweight in Baltimore: A Multilevel Analysis
Patricia O'Campo, PhD, Xiaonan Xue, PhD, Mei-Cheng Wang, PhD, and Margaret O'Brien Caughy, ScD
Introduction Low birthweight remains a significant public health problem in the United States. Low birthweight (defined as a weight of less than 2500 g at birth for a live-bom infant) has declined very little over the past several decades and is even on the rise among some high-risk groups. ' Much research has focused on individuallevel risk factors for low birthweight; individual-level models, however, have been able to explain only a small proportion of the overall variability seen for birthweight.2 Moreover, it is increasingly being recognized that environmental factors contribute to the risk of low birth-
weight.3-7 Previous studies of low birthweight have been limited to individual factors in their conceptualization of social risk.89 Social risk, however, should also include environmental stressors, which shape individual vulnerability and resistance to risk factors for health.40-20 Analyses that include both individuallevel and macrolevel data-referred to as contextual or multilevel models-have several advantages.21 First, multilevel analytic methods are more consistent with social theories than are traditional methods of analysis (e.g., ordinary regression) in that they explicitly accommodate the multiple levels of data.22 Second, multilevel methods can contribute new knowledge to our current understanding of public health issues by allowing for the inclusion of macrolevel factors in our
current explanatory models, thereby bridging the micro-macro gap by increasing our understanding of how contextual factors translate into differences in individual-level risk.23-25 Further, these methods may eliminate potential confounding of individual-level explanatory models
due to the omission of macrolevel fac-
tors.23'26 Finally, by improving our understanding of how contextual factors influence individual health outcomes, we will be better equipped to design effective intervention strategies.'9'20 The analysis presented here is part of a larger study looking at indicators of neighborhood and social risk for poor pregnancy outcome in Baltimore, Md. We were specifically interested in answering the following questions about multilevel analyses and the study of risk factors for low birthweight: (1) Are neighborhoodlevel variables directly related to an increase or decrease in risk of low birthweight? (2) Do individual-level risk factors for low birthweight behave differently depending on the characteristics of the neighborhood in which a woman resides? (3) What are the intervention design or policy implications of including macrolevel variables in research on low
birthweight?
Methods Data Individual-level variables. Computerized birth certificates with nonmissing information on birthweight and matemal characteristics were obtained for the Patricia O'Campo and Margaret O'Brien Caughy are with the Department of Maternal and Child Health, and Mei-Cheng Wang is with the Department of Biostatistics, Johns Hopkins School of Hygiene and Public Health, Baltimore, Md. Xiaonan Xue is with the Department of Biometry and Statistics, School of Public Health, New York State University, Albany. Requests for reprints should be sent to Patricia O'Campo, PhD, Department of Maternal and Child Health, 624 N Broadway, Room 189, Baltimore, MD 21205. This paper was accepted September 16, 1996.
American Journal of Public Health 1113
O'Campo et al.
TABLE 1-Variables Used in Multilevel Analysis of Risk for Low Birthweight, TABLE 1 -Variables Used in Multilevel Analysis of Risk for Low Birthweight, Baltimore, Md, 1985 through 1989 Variable Individual-level Low birthweighta Maternal age, y Maternal education, y Trimester of prenatal care initiationb Health insurance statusc Census tract-level Ratio of home owners to renters No. of community groups Unemployment rate, % Rate of housing violations, % Per capita crime rate, % Average wealth, $000 Per capita income, $000
Type
Distribution, Mean (SD)
Binary Continuous Continuous Categorical Binary
.12 (.32) 24.2 (5.7) 12.0 (3.5) Median is 2 .27 (.44)
Continuous Continuous Continuous Continuous Continuous Continuous log Dummy
1.4 (1.3) 3.0 (2.7) 7 (4.3) 9 (7) 11 (18) 99 (53) 12 (7)
a